BP改进算法在卧龙自然保护区地质灾害危险性评价中的应用研究

发布时间：2018-05-04 03:29

本文选题：人工神经网络 + 地质灾害危险性评价　；参考：《成都理工大学》2017年硕士论文

【摘要】：卧龙自然保护区是我国国家级第三大自然保护区,保护区内珍稀动植物丰富,特别是我国特有珍稀动物大熊猫。卧龙自然保护区位于龙门山中南段,是四川盆地向川西高原的过渡带。2008年汶川地震和2013年芦山地震对卧龙自然保护区造成严重破坏。卧龙自然保护区内地质灾害频发,且对人民生命财产安全以及动植物生态环境造成严重威胁。所以,对卧龙自然保护区地质灾害进行危险性评价,对保护区内展开地质灾害防治工作具有重要意义。本文依托中国科学院数字地球重点实验室开放基金项目“以遥感为主要手段的卧龙大熊猫自然保护区自然灾害与遗产地生境评价”,以及成都理工大学空间信息技术研究所数字地球技术平台进行课题研究。本文通过对研究区域Landsat8遥感图像进行数字图像处理,完成卧龙自然保护区地质灾害解译后提取有效地质灾害信息。并对基础地理信息进行数字化,建立卧龙自然保护区空间信息数据库。根据研究区域特点本文选取海拔高度、地形坡度、地形坡向、地层岩性、河流水系、人类工程活动作为卧龙自然保护区地质灾害危险性评价指标。并基于人工神经网络结构建立卧龙自然保护区地质灾害危险性评价模型,对卧龙自然保护区地质灾害进行危险性评价,得到卧龙自然保护区地质灾害危险性区划图。本文针对标准BP神经网络存在的缺陷,以卧龙自然保护区为例,建立基于附加动量改进算法的地质灾害危险性评价模型、基于自适应调整参数改进算法的地质灾害危险性评价模型、基于共轭梯度改进算法的地质灾害危险性评价模型、基于Levenberg-Marquardt改进算法的地质灾害危险性评价模型。对比地质灾害危险性评价结果发现,四种模型均训练稳定且达到一定精度,基于四种模型下的评价结果相似性较强,危险性分区结果大致相似。但同时在训练时间、迭代次数、危险性区域面积及分布存在细微区别。本文提出对四种基于BP改进算法的地质灾害危险性评价模型下的评价结果进行综合解释,得到卧龙自然保护区地质灾害危险性评价的综合评价结果。该评价结果显示与基于四种改进算法模型下的评价结果相关性较好,且评价结果适用性较好。且基于共轭梯度改进算法的神经网络模型在卧龙自然保护区地质灾害危险性评价中较为适用。为卧龙地质灾害防治工作的规划、部署提供理论支撑。
[Abstract]:Wolong nature reserve is the third Nature Reserve in China. The rare animals and plants are rich in the protected area, especially the rare animal pandas in China. The Wolong nature reserve is located in the southern section of the Longmen mountain. It is the Wenchuan earthquake of the Sichuan basin to the West Sichuan Plateau in.2008 and the Lushan earthquake in 2013 to the Wolong nature reserve. The geological disasters in Wolong natural reserve are frequent, and it has serious threat to the safety of people's life and property and the ecological environment of animals and plants. Therefore, it is of great significance to evaluate the geological hazards in the Wolong nature reserve and to prevent and control the geological disasters in the protected areas. The open fund project of the Key Laboratory of the earth is "the assessment of natural disasters and heritage sites in Wolong Giant Panda Nature Reserve, which is the main means of remote sensing", and the digital earth technology platform of the Institute of space information technology of Chengdu University of Technology. This paper carries out digital images of the Landsat8 remote sensing image of the research area. After the interpretation of geological disaster in Wolong nature reserve, the effective geological disaster information is extracted. The basic geographic information is digitized and the spatial information database of the Wolong nature reserve is set up. According to the characteristics of the study area, the altitude, terrain slope, topographic slope, stratigraphic lithology, river system and human engineering activities are selected. It is the risk assessment index of geological hazard in Wolong nature reserve. Based on the artificial neural network structure, the risk assessment model of geological hazard in Wolong nature reserve is established. The hazard assessment of geological disasters in Wolong nature reserve is evaluated, and the geological hazard zoning map of the Wolong natural reserve is obtained. This paper aims at the standard BP neural network. The existing defects of the collaterals, taking the Wolong nature reserve as an example, set up a geological hazard assessment model based on the additional momentum improvement algorithm, based on the adaptive adjustment parameter improvement of the geological hazard risk assessment model, based on the conjugate gradient improved algorithm of geological hazard risk assessment model, based on the Levenberg-Marquardt improvement. The result of the evaluation of geological hazard risk assessment shows that the four models are both trained and achieved a certain precision. The results of the evaluation results based on the four models are more similar, and the results of the hazard zoning are roughly similar, but at the same time, the training time, the number of iterations, the area and distribution of the danger area and the distribution exist. This paper gives a comprehensive interpretation of the evaluation results of four geological hazard assessment models based on the BP improved algorithm, and obtains the comprehensive evaluation results of the geological hazard assessment of the Wolong nature reserve. The results show that the results are well correlated with the evaluation results based on the four improved algorithm models, and the evaluation results are good. The applicability of the price results is better. And the neural network model based on the conjugate gradient improvement algorithm is more suitable for the evaluation of geological hazard in Wolong nature reserve. It provides theoretical support for the planning of Wolong geological disaster prevention and control work.

【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P694

【参考文献】